Enjoying this issue?
Get tomorrow's AI & engineering digest in your inbox — hand-picked, summarized, and always spam-free.
TLDR
QAC is a local AI SDK that lets you install and run multiple AI models (speech-to-text, embeddings, LLM, text-to-speech) with a single npm install, eliminating the need to configure separate runtimes. The platform supports over a dozen capabilities including RAG, image/video generation, and fine-tuning, and all models run locally as files on your machine.
Key points
- QAC by Tether provides a single npm install to set up a complete local AI system with multiple model types.
- The SDK automatically downloads and runs models like Whisper (speech-to-text), Gemma (embeddings), Qwen 3 (LLM), and Supersonic (text-to-speech).
- Users avoid the hassle of configuring separate runtimes for each model type (e.g., Whisper, Ollama).
- QAC ships with over a dozen capabilities out of the box, including LLM, fine-tuning, RAG, image/video generation, and transcription.
- Running models locally as files means no rate limits, no repricing, and no deprecation risks.
- The same API works across all operating systems.
- A demo showed a voice assistant using Whisper, Gemma for RAG, Qwen 3 for answer generation, and Supersonic for speech output.
Tools mentioned
Techniques
- Retrieval-Augmented Generation (RAG)
- Fine-tuning
- Model serving via SDK
Stop scrolling. Start reading smarter.
Receive the day's most important AI & engineering updates in one concise email. No spam.
Transcript (captions)
Take a look at this. I have four different kinds of models on my laptop. And I brought all of this onto my machine with a single npm install, which is fantastic because normally when you have these different kinds of models, it's four different run times, right? You have something for your speech to text, for your embeddings, for your LLM, and your text to speech like using Whisper and Olama for example. But it's annoying to have to configure all those different services just to have your local AI set up.
And that single mpm install is QAC by Tether. The ultimate local AI SDK. It gives you the complete suite for running anything you would ever need for local AI within a single platform. Just take a look at all the different AI capabilities that are available to you out of the box. LLM, fine-tuning, rag, image generation, even video generation, transcription.
Of course, we have our speechto text and texttospech models. Literally anything that you're going to be doing with AI, you have it within the ecosystem here. So, going back to the little demo that I have for you here, we can take a look at the specific models that we're using. And so, I'm just importing these from the SDK. It automatically downloads and runs them for me.
So, I'm using Whisper for my speechtoext. For embedding, I'm using Gemma. Then, my LLM is Quen 3. And then for my texttospech, Supersonic. So, all really popular models.
But the beauty of this is I don't have to configure each one individually. I just import them and they're good to go. So, let me show you the full demo here really quickly. Now, keep in mind this is not supposed to be a fast voice assistant. It's just a good use case to show you a lot of the different kinds of models we can bring onto our system.
And so, I'm going to run this here and it's going to tell me to ask a question like what is the database that we ended up going with? And then it's first going to use Whisper here to transcribe that to text. And then we're using the embedding model to find notes. So we're performing rag. And then we have the text generation.
So calling upon Quen to give us the final answer. And then we'll hear it in just a second here. We'll hear the answer from the model. >> We chose Postgres over for reporting queries. >> All right, there we go.
And that's it. So quite a simple example here, but it's pretty awesome how easy it was to bring in these models and build out this full pipeline. Everything running locally. QAC has over a dozen capabilities that ship out of the box and the same API works on every operating system. And since all these models are just files on my computer, I have no risk of being rate limited, repriced, or having things deprecated from under me.
Thanks to QAC for working with me on this video, and they would love it if you go and give them a star on GitHub. I'll have a link to their GitHub in the description.